Vous êtes sur la page 1sur 9

Global Ecology and Biogeography, (Global Ecol. Biogeogr.

) (2017)

RESEARCH Global exposure of carnivores


PA P E R
to roads
Ana Ceia-Hasse1,2,3,4*, Lus Borda-de-Agua
 3,4
, Clara Grilo5 and
Henrique M. Pereira1,2,3,4

1
German Centre for Integrative Biodiversity ABSTRACT
Research (iDiv) Halle-Jena-Leipzig, 04103
Aim Land-use change is a major threat to biodiversity globally. Roads cause
Leipzig, Germany, 2Institute of Biology,
Martin Luther University Halle-Wittenberg, direct mortality and limitation of individual movements, which may isolate
06108 Halle (Saale), Germany, populations and affect their viability in the long term. Here we provide the
3
Infraestruturas de Portugal Biodiversity first comprehensive global assessment of the exposure of terrestrial mammalian
Chair, CIBIO/InBIO, Centro de Investigac~ao carnivores to roads using an integrated modelling framework.
em Biodiversidade e Recursos Gene ticos,
Location Global.
Campus Agra rio de Vair~ao, Universidade do
Porto, 4485-661, Vair~ao, Portugal, 4CEABN/ Methods We estimated critical road densities and critical patch sizes for each
InBio, Centro de Ecologia Aplicada species based on a spatially explicit model and life-history traits. We calculated the
Professor Baeta Neves, Instituto Superior distribution of landscape fragment sizes for each carnivore species by intersecting
de Agronomia, Universidade de Lisboa, global road density with each species range. The proportion of a species
Tapada da Ajuda, 1349-017 Lisboa,
geographical range with fragments below the critical patch size is used as an index of
Portugal, 5Centro Brasileiro de Estudos em
the vulnerability to roads.
Ecologia de Estradas (CBEE), Setor de
Ecologia Aplicada, Departamento de Results We found that the carnivores expected to be most exposed to roads belong
Biologia, Universidade Federal de Lavras, to families Felidae, Ursidae, Mustelidae, Canidae and Procyonidae. Approximately
Campus Universitario, 37200-000 Lavras, one-third of the species most affected have not been identified by the IUCN as
Minas Gerais, Brasil threatened by roads. Our model projects time to extinction that may be as low as
one century for some species, such as the endangered Iberian lynx. Species are
expected to be more exposed in areas with medium to high road density but,
surprisingly, also in areas where road density is relatively low. Hotspots of the
number of species locally endangered by roads occur in North America and Asia.
Main conclusions Our results suggest the need to reassess the status and
threats of those species that have not been previously recognized as strongly
*Correspondence: Ana Ceia-Hasse, German
Centre for Integrative Biodiversity Research affected by roads. Our framework can be applied at different spatial scales, to
(iDiv) Halle-Jena-Leipzig, Deutscher Platz 5e, assess the effects of the development of the road network and inform
04103 Leipzig, Germany. prioritization schemes for road building, and to identify areas for conservation,
E-mail: ana.ceia_hasse@idiv.de
This is an open access article under the terms
and species requiring particular mitigation and restoration measures.
of the Creative Commons Attribution License,
Keywords
which permits use, distribution and
reproduction in any medium, provided the Biodiversity conservation, carnivores, dispersal, global assessment, mortality,
original work is properly cited. population viability, road impact.

INTRODUCTION
thus affecting their viability (van der Ree et al., 2015). The current
Land-use change is one of the major drivers of biodiversity loss and the projected expansion of the road network pose threats that
(Pereira et al., 2012). Biodiversity is projected to continue to should be evaluated across scales in order to minimize their nega-
decline despite societys increasing efforts to decelerate this trend tive effects (Laurance et al., 2014) and to ensure the long-term via-
(Tittensor et al., 2014). The road infrastructure is a cause of mor- bility of populations. The effect of roads on animal populations
tality for many species and is also a major barrier to movement. has been addressed by various studies at local and regional
Roads can subdivide populations and reduce population sizes, scales, focusing on components such as behavioural and

C 2017 The Authors. Global Ecology and Biogeography published


V DOI: 10.1111/geb.12564
by John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/geb 1
A. Ceia-Hasse et al.

physiological responses to roads (Grilo et al., 2012; Navarro- 


rate (Borda-de-Agua et al., 2011) (see model description in
Castilla et al., 2014), the effect of road noise and pollution Methods).
(Hopkins et al., 2013; McClure et al., 2013) and projections of This work is the first where a mechanistic population
population declines and loss of genetic diversity (for a review model has been combined with life-history data, biogeo-
see Balkenhol & Waits, 2009; Jackson & Fahrig, 2011). How- graphical data and land-use data to produce a global assess-
ever, studies assessing the impact of roads on population per- ment of population viability for a complete taxonomic
sistence remain uncommon, and upscale, at best, to national group. Despite not taking into account other factors that
levels, focusing on a limited number of species (Beaudry et al., influence how populations are affected by roads, such as the
2008; Borda-de-Agua et al., 2014). Assessments on larger behaviour of animals towards roads, the type of road or traf-
scales up to global, and across species, have never been con- fic intensity, or indirect mortality caused by resource inacces-
ducted (but see Torres et al., 2016). sibility and population subdivision (Jaeger & Fahrig, 2004;
Here, we present a new spatially explicit modelling frame- Jaeger et al., 2005), our work opens the way for similar
work to assess the exposure of biodiversity to a major threat global mechanistic assessments of other threats. Our frame-
the road infrastructure and to map hotspots of road work can be applied at different spatial scales, promoting
impact on biodiversity globally. We apply this framework to quantitative-based assessments directing road development
a particularly vulnerable group: terrestrial carnivore species. (or avoidance) and reducing conflicts with biodiversity
We carry out an assessment of which terrestrial carnivore conservation.
species are more affected by roads at the global level, as well
as where within their range they are most at risk. Mamma- METHODS
lian carnivores usually have life-history traits that make them
particularly vulnerable to the effects of human-induced envi- For each species analysed, we determined its intrinsic vulner-
ronmental changes, such as relatively high mobility and low ability to roads by computing two metrics: (1) the maximum
reproductive rates (Grilo et al., 2015). Moreover, they are road density above which populations are expected to go
important for maintaining ecosystem function, structure and locally extinct (Dmax) and (2) the minimum size of favour-
resilience (Ripple et al., 2014). able habitat patches that are delimited by roads, below which
We use a simple, spatially explicit reactiondiffusion populations are expected to go locally extinct (Amin) (Borda-

de-Agua et al., 2011). We computed these metrics using
demographic model describing population dynamics and the
dispersal of individuals (Skellam, 1951; see also Cantrell & species-specific empirical data on life-history traits (see
Cosner, 2003). Skellams (1951) model can be used to simu- details below; a list of the data sources is found in the
late sourcesink dynamics (Pulliam, 1988) of populations Appendix). We then calculated the observed road density
occupying several habitats within a landscape, where the (Dobs), and the sizes of the patches delimited by roads (Aobs),
population growth rate can be positive in favourable habitats that exist within each species range, to compare with the
(source habitats) or negative in unfavourable habitats (sink critical values of road density and of patch size, respectively.
habitats) (Pereira & Borda-de-Agua, 2013). It has been used We applied this approach to 232 terrestrial carnivore species
to assess species vulnerability to land-use change (Pereira for which an IUCN range map is available (IUCN, 2015),
et al., 2004; Pereira & Daily, 2006) and to develop metrics to but excluded those species that can disperse through water
analyse the impact of road networks on population persist- (Tables S1 & S2 in the Supporting Information).

ence (Borda-de-Agua et al., 2011).
 Critical values of Dmax and of Amin
The analysis performed by Borda-de-Agua et al. (2011)
considered a population occupying a landscape composed of The expressions for Dmax and Amin were derived by Borda-
favourable habitat patches, where the population growth rate 
de-Agua et al. (2011) considering a population in a land-
is positive, surrounded by roads, which are unfavourable scape composed of patches surrounded by roads. The
habitat where the population growth rate is negative. The patches consist of favourable habitat where the population
model predicts how road density and the size of favourable growth rate is positive and equal to the intrinsic popula-
habitat patches that are delimited by roads determine popu- tion growth rate (r1), and the roads consist of unfavourable
lation viability in landscapes fragmented by roads. In particu- habitats where the growth rate is negative (r0). The disper-
lar, it shows that the minimum patch size required for a sal distance of individuals is modelled by its dispersal var-
population to persist increases with population mean disper- iance (r2). The dynamics are given by (Skellam, 1951;
sal distance, and decreases with intrinsic population growth 
Borda-de-Agua et al., 2011):

8 2  
>
> r N x; y; t
> r2 N x; y; t 1r1 N x; y; t 12
< if x; y 2
= road
dN x; y; t 2 K
5 ;
dt >
> 2
> r r2 N x; y; t 1r N x; y; t
: if x; y 2 road
0
2

2 Global Ecology and Biogeography, V


C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd


Exposure of carnivores to roads

where N(x, y, t) is the population density at location (x, y) at Pereira & Daily (2006), and assuming: (1) a constant mortal-
time t, K is the carrying capacity and r2 is d2/dx2 1 d2/dy2. ity rate, l; (2) zero fecundity before the breeding age, b; (3)
The first term on the right-hand side of the equation (top fecundity equals the number of female offspring (50% of the
and bottom branches) describes the changes in population litter size), b, at regular birth pulse intervals after the breed-
density in space and time on the basis of the dispersal dis- ing age; and (4) birth pulse intervals are spaced by the mean
tance, which is assumed to follow a Gaussian distribution. interval between litters, D. The implicit equation for r1 is
The second term describes logistic growth outside roads (top then:
branch) and the population decay (since r0 is assumed to be 1 X
1
negative) on roads (bottom branch) (Borda-de-Agua  et al., b3 dx2yD2be2r1 1lx dx51;
2011). While r0 mathematically is the population growth rate 0 y50
on roads, here it is a measure of the loss of individuals from
the population (i.e. roads behave as a sink habitat), repre- where d(x) is the birth pulse function, which has a value of
senting the instantaneous mortality rate when an animal 1/T for x between 0 and T and 0 elsewhere. This equation
crosses a road. can be solved numerically to determine r1.
We estimated the rate of population decay on roads, r0, as:
Maximum road density and minimum patch size

The expressions for the maximum road density above which r0 5 2l 3103 year21 ;
populations cannot persist (Dmax) and for the minimum
patch size below which populations cannot persist (Amin) where l is the annual natural mortality rate (see below). We
were obtained by solving the equation above for simplified assume a 1000 times higher mortality rate (and zero birth
rate) on roads than the natural mortality rate. For example,
situations, providing easy rules of thumb to estimate popula-
for a species with natural mortality rate of 0.5 year21, this
tion viability in landscapes fragmented by roads. See Borda-
 corresponds to a mortality rate per crossing of 500 year21 on
de-Agua et al. (2011) and its corresponding Supporting
roads. We assessed the robustness of this assumption by com-
Information for the details on the derivation of Dmax and of
paring the results with r0 5 l 3 102, and with r0 5 l 3
Amin. Dmax was derived assuming very large dispersal (r2 !
104 (Table S3). Our estimates of Dmax are consistent with
1) and large carrying capacity (K ! 1, so the term 1
previous maximum road density estimates (e.g. Anderson
N(x, y, t)/K in the top branch of the equation above is not
et al., 2011).
considered), and ignoring the spatial location of the roads
Dispersal variance (r2) was computed assuming Gaussian
and considering road density only:
dispersal as:
Dmax 5 r1 =r1 1 jr0 j;
r2 5 rm =1:182 3l;
where r1 is the growth rate of the population and r0 is mor-

tality on roads. In Borda-de-Agua et al. (2011) the maximum where rm is the dispersal median and l is the annual mor-
road density was computed as a ratio of two areas, and is tality rate, which converts the dispersal median from km2
therefore a dimensionless quantity. Here we divided Dmax by generation21 to km2 year21 (Pereira & Daily, 2006).
We computed these parameters (r1, r0 and r2) for each
0.01 to compare it with the road density that is observed
species using species-specific life-history data (Tables S1 &
within each species range (considering a constant road width
S2). We computed mortality rate (l) as the inverse of mean
of 0.01 km; Dobs in km km22; see below).
life span, and the dispersal median (rm) from home range
The expression for the minimum patch size below which
(HR) data as rm 5 7HR (Bowman et al., 2002). Since not
populations go extinct (Amin) was derived assuming infinite
all data were available for all species, we established allomet-
carrying capacity (K ! 1), and that individuals always die
ric relationships (Table S4) based on the available data, and
when crossing a road (r0 ! 1), considering the location of
then used these relationships to estimate the missing life-
the roads explicitly and assuming their configuration to be
 history values. When data for body mass and litter size were
that of a square grid (Borda-de-Agua et al., 2011):
lacking for a given species we used the mean of the genus, or
Amin 5p2 r2 =r1 ; the mean of the corresponding family (Pereira & Daily,
2006). The percentage of missing life-history values ranged
where r2 is dispersal variance and r1 is the growth rate of from 12.9% (for body mass) to 92.7% (for mortality rate).
the population.
Observed values of road density (Dobs) and of patch
Population parameters: growth rates and dispersal size (Aobs)
To compute Dmax and Amin we first estimated the parameters After computing the critical values of road density and patch
of the model (r1, r0 and r2) for each species. We computed size, we assessed what is observed within each species range
the intrinsic population growth rate, r1, using a simplified with respect to roads. We did this by intersecting each species
version of the Euler equation following the approach by range map from the IUCN (IUCN, 2015) with the road
Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd 3


A. Ceia-Hasse et al.

Figure 1 Observed and critical values of patch size and of road density for two species: frequency of observed patch sizes (Aobs) and the
relation with minimum patch size (Amin) (a), and the ratio of theoretical maximum road density to observed road density (Dmax/Dobs)
(b). The puma (Puma concolor (Linnaeus, 1771)) is among the 5% of species with lower P[Aobs > Amin] and the 5% of species with
smaller Dmax/Dobs. In contrast, the black slender mongoose (Herpestes flavescens Bocage, 1889) has high P[Aobs > Amin] and high Dmax/
Dobs. Fewer patches larger than Amin (areas shaded in grey in (a)), or with a small value of Dmax/Dobs (in (b)), reveal a highly vulnerable
species.

network from Open Street Map (OSM) (Geofabrik, 2015), combines the areas of all patches within the region investi-
using QGIS v.2.2.0 (QGIS Development Team, 2014). The gated, as well as its total area, into one metric of landscape
road categories included from the OSM data were: motor- fragmentation; in our model each Aobs is simply the area of
way, trunk, primary, secondary, tertiary, minor, road one patch (see below for the use of Aobs in ranking species
and unclassified. exposure to the road network).
We then computed the observed road density (Dobs) as the
ratio of the road length that exists within the species range
Species exposure to the road network
to the species range area. We also computed the area of each
patch that is delimited by roads (observed patch sizes, Aobs), We ranked the species with regard to their exposure to the
existing within each species range. This differs from the effec- extant road network by computing for each species: (1) the
tive mesh size metric proposed by Jaeger (2000), which ratio of the maximum road density to the observed road
4 Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd


Exposure of carnivores to roads

Table 1 Species most exposed to roads.

Smaller Smaller Percentage of Threatened by Red List


Common name Species Family Dmax/Dobs P[Aobs > Amin] range affected roads IUCN status

Iberian lynx Lynx pardinus Feli 1 1 100 1 EN


Japanese badger Meles anakuma Must 1 1 100 LC
Japanese marten Martes melampus Must 1 1 100 LC
Bobcat Lynx rufus Feli 1 93 1 LC
Stone marten Martes foina Must 1 1 90 LC
Sloth bear Melursus ursinus Ursi 1 89 1 VU
Nilgiri marten Martes gwatkinsii Must 1 1 86 1 VU
Asiatic black bear Ursus thibetanus Ursi 1 86 1 VU
Puma Puma concolor Feli 1 1 69 1 LC
American black bear Ursus americanus Ursi 1 1 68 1 LC
Darwins fox Pseudalopex fulvipes Cani 1 67 CR
Brown bear Ursus arctos Ursi 1 65 1 LC
Jaguarundi Herpailurus yagouaroundi Feli 1 60 1 LC
Leopard Panthera pardus Feli 1 59 1 NT
Pygmy racoon Procyon pygmaeus Procy 1 50 1 CR
Coyote Canis latrans Cani 1 48 LC
Jaguar Panthera onca Feli 1 38 1 NT

These are the species within the 5% lowest values of Dmax/Dobs or within the 5% lowest values of P[Aobs > Amin]. We indicate the percentage of
100 km 3 100 km grid cells where each species is present having Dobs > Dmax or at least one patch with Aobs smaller than Amin, whether the spe-
cies is identified by the IUCN as being threatened by roads, and the species IUCN Red List status: CR, Critically Endangered; EN, Endangered;
VU, Vulnerable; NT, Near Threatened; LC, Least Concern. Family names are truncated to the first four letters: Cani, Canidae; Feli, Felidae; Must,
Mustelidae; Proc, Procyonidae; Ursi, Ursidae.

density (Dmax/Dobs), and (2) the proportion of patches 1


Text 5   :
delimited by roads observed within the species range with r2 p2
Aobs 2r1
area (Aobs) larger than the minimum patch size (Amin),
P[Aobs > Amin] (Fig. 1). Then for each of these quantities, we Time to extinction here is the time that it takes for the spe-
selected the species within the lower fifth percentile as the cies to disappear from 90% of the area where Aobs is smaller
species that are expected to be more exposed to roads. In than Amin, and Aobs is the area of the largest patch smaller
this way, those species that are expected to be most exposed than Amin within that area.
to roads are the ones with a lower Dmax/Dobs and/or the ones
with a smaller P[Aobs > Amin].
RESULTS
For each species in the lower fifth percentile we also
mapped where it is expected to be more exposed to roads Using the fifth percentile criterion, we identified 17 species as
within its range. We did so using 100 km 3 100 km grid those most exposed to roads. They belong to the families
cells and identifying in which cells Dobs is higher than Dmax, Felidae (six species, corresponding to 17% of the species ana-
or there is at least one patch with Aobs smaller than Amin for lysed in this family), Ursidae (four species, 57%), Mustelidae
population persistence. This is a conservative estimate, (four species, 9%), Canidae (two species, 6%) and Procyoni-
because it does not imply that a species is more exposed to dae (one species, 7%) (Table 1). The percentage of the range
roads in all the area of a grid cell, but only in parts of that where each species is expected to be affected varies from
grid cell (where Aobs < Amin). 38% to 100% (Table 1). These are the parts of the range
where each species is projected to disappear or have low
abundance. The Iberian lynx (Lynx pardinus (Temminck,
Time to extinction
1827)), an endangered species endemic to the Iberian Penin-

Borda-de-Agua et al. (2011) also derived an expression for sula, is estimated to go extinct from 90% of the area where
computing the time to extinction of a population in a patch Aobs is smaller than Amin in 114 years.
with area smaller than the minimum patch size (Amin). Using Among the species expected to be most exposed to roads, 71%
this expression, we determined the time to extinction (Text) (12 species) are also classified by the IUCN as threatened by roads
for the species within the lower fifth percentile of (Table 1). However, we also identified species for which roads are
P[Aobs > Amin] for which there is at least one patch with Aobs not listed as a threat by the IUCN (IUCN, 2015), some of which
smaller than Amin in all the grid cells where those species are are expected to be affected in more than 75% of their range.

present, as follows (Borda-de-Agua et al., 2011): These include species that have been categorized as of Least
Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd 5


A. Ceia-Hasse et al.

Figure 2 Road density (km km22) per 100 km 3 100 km grid cell. We intersected the global road network (Geofabrik, 2015) with the world
map in 100 km 3 100 km grid cells and then summed the road length in each grid cell. The scale represents 50 road density classes obtained
using the Jenks natural breaks optimization method (QGIS Development Team, 2014).

Concern by the IUCN, and for which we have identified another DISCUSSION
factor of threat: the stone marten (Martes foina (Erxleben,
We have assessed the impact of roads on terrestrial carnivore
1777)), which despite being widespread is vulnerable to traffic
species at the global level, using a species-specific, spatially
and is frequently road-killed (Grilo et al., 2009), and the Japanese
explicit approach thereby identifying not only species that
badger (Meles anakuma Temminck, 1844) and Japanese marten
are expected to be most exposed to roads at the global level,
(Martes melampus (Wagner, 1840)), which both occur only in
but also where, within its range, each species is expected to
Japan and are estimated to disappear from 90% of the areas
be more affected.
where Aobs is smaller than Amin in 9 and 17 years, respectively. Our approach has several limitations that need to be con-
When using the lower 25th percentile as a criterion to
sidered when interpreting the results of our analysis. The
select the species that are expected to be most exposed to maximum road density (Dmax) was derived assuming unlim-
roads, we captured 39 out of the 53 species (74%) that are ited dispersal ability, and the minimum size of favourable
listed by the IUCN as threatened by roads (among the 232 habitat patches that are surrounded by roads (Amin) was
species analysed; Table S2). However, we also identify 50 derived assuming that individuals always die when crossing a
more species besides the species already identified at the road (see Methods and Borda-de-Agua  et al., 2011). Such
lower fifth percentile (Table 1) for which roads are not assumptions are simplifications of real ecological systems:
listed as a threat by the IUCN. these metrics may underestimate the maximum possible road
Species are clearly affected in regions with a medium to density and overestimate the minimum possible patch size
high road density, as in Europe (mean road density 6 SD necessary for population persistence, and it is possible that
0.65 6 0.54 km km22), North America (eastern USA, mean 6 even if Dobs > Dmax, or when Aobs < Amin, populations persist
SD 0.49 6 0.24 km km22; south-central Canada, 0.48 6 
(Borda-de-Agua et al., 2011). Moreover, the behaviour of
0.31 km km22) and Japan (mean 6 SD 0.43 6 0.32 km individuals towards roads influences how roads affect species
km22), but also in regions with relatively low road density, (Jaeger & Fahrig, 2004; Jaeger et al., 2005). Many large pred-
such as Africa (mean 6 SD 0.04 6 0.07 km km22) (Figs 2 & ator species move along low-traffic roads (Forman &
3). In Africa, only one of the fifth percentile-selected species Alexander, 1998); for example, wolves can select low-use
occurs, but the leopard (Panthera pardus (Linnaeus, 1758)), roads as travel routes (e.g. Whittington et al., 2005). Also,
having low Dmax and large Amin (Table S2), shows a strong road mortality may be compensated for by increased fecun-
effect of roads throughout its range. The highest number of dity or survival (e.g. Seiler & Helldin, 2006).
species affected by roads per 100 km 3 100 km grid cell is For these reasons, it is not appropriate to use Dmax or
observed in North America (up to six species), followed by Amin as strict thresholds for population persistence. Instead,
Asia (four species), South America (three species) and we have ranked the species analysed in relation to one
Europe (two species) (Fig. 3). We observe a similar pattern another to detect which species would be more exposed to
when we use the 25th percentile (Fig. S1). roads than other species, and we have estimated in which
6 Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd


Exposure of carnivores to roads

Figure 3 The number of species expected to be more exposed to the impact of roads per 100 km 3 100 km grid cell. The species
mapped are within the lower fifth percentile of Dmax/Dobs or of P[Aobs > Amin]. The scale represents the total number of species per grid
cell for which Dobs is higher than Dmax, or Aobs is smaller than Amin.

areas within its range a species would be more exposed to because it would have been impractical in an analysis per-
roads than in other areas of its range. We did so because formed at the global scale and for such a large number of
even though populations may persist where Dobs > Dmax or species. Assessments performed at smaller spatial scales, or
where Aobs < Amin they are expected to be more exposed to for a smaller number of species, could include computing
the impact of roads in those areas than in areas where separate Dmax and separate Dobs for different road types.
Dobs < Dmax, or Aobs > Amin, respectively. Our results suggest the need to reassess the status and
Despite its limitations, our framework provides a way to threats of those species that have not been previously recog-
obtain a first approximation to pinpoint species that may be nized as exposed to roads. In such cases applying a modelling
more exposed to the impact of roads, to identify the areas approach such as ours, which explicitly combines extrinsic
within a species range where it may be more exposed to factors of threat and intrinsic species traits, will allow us to
roads and also to highlight regions that may have a higher move from a descriptive to a mechanistic, biologically sound
number of species more exposed to roads. evaluation of threats (Lee & Jetz, 2010; Dirzo et al., 2014).
Time to extinction should also be interpreted with caution Our approach brings together process-based models that
given the assumptions made when deriving the expressions link species-specific life histories, population dynamics and
for Amin and Text (see Methods and Borda-de-Agua  et al., dispersal with biogeographical and land-use data in this
2011) and the unknown effects of the behaviour of each spe- case, the road infrastructure. However, as discussed above,
cies. While our approach allows us to estimate the time to other factors that we did not account for can influence how
extinction in those areas within a species range where roads affect animals. In more localized studies it may be
Aobs < Amin, the assumption that animals always die when desirable to incorporate those issues in order to obtain a
crossing a road is a simplification. Moreover, the impact of refined understanding of the impacts across species and
roads on population persistence, and therefore time to regions, but for that it is necessary to deepen our knowledge
extinction, will also depend on their sensitivity to the effects on the impact of roads, and strengthen the link between
of roads (including not only road mortality but also habitat empirical knowledge and models (e.g. using models to guide
loss, resource inaccessibility and population subdivision), on data collection, or collecting parameters that are implement-
the type of behavioural response of the animals towards able for models).
roads, on the types of roads, on traffic intensity (Jaeger Our analysis differs from a simple spatial overlap between
et al., 2005), on habitat affinities and on the variation of species ranges and road density. Despite its limitations, it
population density within species ranges (Grilo et al., 2014). provides a first quantitative approximation to assess whether
Because these factors were not explicitly taken into account the road density that is observed within a species range may
in our analysis, Text should be considered as a first be beyond a threat threshold. By integrating species traits,
approximation. this analysis also allowed us to detect species which, despite
We also grouped all roads together when computing Dobs, having relatively low road density within their range, are
and we did not differentiate road mortality according to the nevertheless estimated to be more exposed to roads than
type of road (e.g. Jaeger et al., 2005; Grilo et al., 2009) other species for which road density is relatively high within
Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd 7


A. Ceia-Hasse et al.

their range. Such inferences would not have been possible Grilo, C., Bissonette, J.A. & Santos-Reis, M. (2009) Spatial
using a simple spatial overlap analysis between species ranges temporal patterns in Mediterranean carnivore road casual-
and road density. ties: consequences for mitigation. Biological Conservation,
By pinpointing where species are expected to be affected 142, 301313.
by roads within their ranges, and which regions have more Grilo, C., Sousa, J., Ascens~ao, F., Matos, H., Leit~ao, I.,
species vulnerable to roads, this type of spatial analysis can Pinheiro, P., Costa, M., Bernardo, J., Reto, D., Lourenco,
be used for identifying areas for conservation, for prioritizing R., Santos-Reis, M. & Revilla, E. (2012) Individual spatial
regions where mitigation measures, such as passages or fen- responses towards roads: implications for mortality risk.
ces, should be implemented and for informing the develop- PLoS One, 7, e43811.
ment of schemes for road building (Laurance et al., 2014). Grilo, C., Reto, D., Filipe, J., Ascens~ao, F. & Revilla, E. (2014)
Moreover, this framework can also be applied at different Understanding the mechanisms behind road effects: linking
spatial and temporal scales, for example for conservation or occurrence with road mortality in owls. Animal Conserva-
management purposes at the local scale, for environmental tion, 17, 555564.
impact assessments or for projecting the effects of future Grilo, C., Smith, D.J. & Klar, N. (2015) Carnivores: strug-
large-scale road network developments, which can then feed, gling for survival in roaded landscapes. Handbook of road
for example, World Bank projections or scenarios for global ecology (ed. by R. van der Ree, D. J. Smith and C. Grilo),
biodiversity change (Pereira et al., 2010). pp. 300312. Wiley, Chichester.
Hopkins, G.R., French, S.S. & Brodie, E.D.J. (2013) Increased
ACKNOWLEDGEMENTS frequency and severity of developmental deformities in
rough-skinned newt (Taricha granulosa) embryos exposed
This study was supported by DFG, PTDC/AAC-AMB/117068/
to road deicing salts (NaCl & MgCl2). Environmental Pollu-
2010, SFRH/BPD/64205/2009, UID/BIA/50027/2013 and
tion, 173, 264269.
POCI-01-0145-FEDER-006821.
IUCN (2015) The IUCN Red List of threatened species. Version
2015.4. Available at: http://www.iucnredlist.org
REFERENCES
Jackson, N.D. & Fahrig, L. (2011) Relative effects of road
Anderson, S.C., Farmer, R.G., Ferretti, F., Houde, A.L.S. & mortality and decreased connectivity on population genetic
Hutchings, J.A. (2011) Correlates of vertebrate extinction diversity. Biological Conservation, 144, 31433148.
risk in Canada. BioScience, 61, 538549. Jaeger, J.A.G. (2000) Landscape division, splitting index, and
Balkenhol, N. & Waits, L.P. (2009) Molecular road ecology: effective mesh size: new measures of landscape fragmenta-
exploring the potential of genetics for investigating trans- tion. Landscape Ecology, 15, 115130.
portation impacts on wildlife. Molecular Ecology, 18, 4151 Jaeger, J.A.G. & Fahrig, L. (2004) Effects of road fencing on
4164. population persistence. Conservation Biology, 18, 16511657.
Beaudry, F., deMaynadier, P.G. & Hunter, M.L.J. (2008) Identi- Jaeger, J.A.G., Bowman, J., Brennan, J., Fahrig, L., Bert, D.,
fying road mortality threat at multiple spatial scales for Bouchard, J., Charbonneau, N., Frank, K., Gruber, B. & von
semi-aquatic turtles. Biological Conservation, 141, 25502563. Toschanowitz, K.T. (2005) Predicting when animal popula-

Borda-de-Agua, L., Navarro, L., Gavinhos, C. & Pereira, tions are at risk from roads: an interactive model of road
H.M. (2011) Spatio-temporal impacts of roads on the per- avoidance behavior. Ecological Modelling, 185, 329348.
sistence of populations: analytic and numerical approaches. Laurance, W.F., Clements, G.R., Sloan, S., OConnell, C.S.,
Landscape Ecology, 26, 253265. Mueller, N.D., Goosem, M., Venter, O., Edwards, D.P., Phalan,

Borda-de-Agua, L., Grilo, C. & Pereira, H.M. (2014) Model- B., Balmford, A., Van Der Ree, R. & Arrea, I.B. (2014) A
ing the impact of road mortality on barn owl (Tyto alba) global strategy for road building. Nature, 513, 229232.
populations using age-structured models. Ecological Model- Lee, T.M. & Jetz, W. (2010) Unravelling the structure of spe-
ling, 276, 2937. cies extinction risk for predictive conservation science. Pro-
Bowman, J., Jaeger, J.A.G. & Fahrig, L. (2002) Dispersal dis- ceedings of the Royal Society B: Biological Sciences, 278,
tance of mammals is proportional to home range size. 13291338.
Ecology, 83, 20492055. McClure, C.J.W., Ware, H.E., Carlisle, J., Kaltenecker, G. &
Cantrell, R.S. & Cosner, C. (2003) Spatial ecology via reac- Barber, J.R. (2013) An experimental investigation into the
tiondiffusion equations. Wiley, Chichester. effects of traffic noise on distributions of birds: avoiding
Dirzo, R., Young, H.S., Galetti, M., Ceballos, G., Isaac, N.J.B. the phantom road. Proceedings of the Royal Society B: Bio-
& Collen, B. (2014) Defaunation in the Anthropocene. Sci- logical Sciences, 280, 20132290.
ence, 345, 401406.  Mata, C., Ruiz-Capillas, P., Palme, R.,
Navarro-Castilla, A.,
Forman, R.T.T. & Alexander, L.E. (1998) Roads and their Malo, J.E. & Barja, I. (2014) Are motorways potential stres-
major ecological effects. Annual Review of Ecology and Sys- sors of roadside wood mice (Apodemus sylvaticus) popula-
tematics, 29, 207231. tions? PLoS One, 9, e91942.
Geofabrik (2015) OpenStreetMap-Shapefiles. Available at: Pereira, H.M. & Borda-de-Agua,  L. (2013) Modelling
http://download.geofabrik.de/ (accessed January 2015). biodiversity dynamics in countryside and native habitats.
8 Global Ecology and Biogeography, V
C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd


Exposure of carnivores to roads

Encyclopedia of biodiversity, pp. 321328. Elsevier, New SUPPORTING INFORMATION


York.
Additional supporting information may be found in the
Pereira, H.M. & Daily, G.C. (2006) Modeling biodiversity
online version of this article at the publishers web-site:
dynamics in countryside landscapes. Ecology, 87, 18771885.
Pereira, H.M., Daily, G.C. & Roughgarden, J. (2004) A Figure S1 Number of species expected to be more exposed
framework for assessing the relative vulnerability of species to the impact of roads per 100 km 3 100 km grid cell,
to land-use change. Ecological Applications, 14, 730742. within the 25th percentile of Dmax/Dobs or of P[Aobs > Amin].
Pereira, H.M., Leadley, P.W., Proenca, V. et al. (2010) Scenar- Table S1 Species analysed and life-history data.
Table S2 Parameters for the species analysed.
ios for global biodiversity in the 21st century. Science, 330,
Table S3 Dmax and Dmax/Dobs computed using r0 5 l 3
14961501.
102, r0 5 l 3 103 and r0 5 l 3 104.
Pereira, H.M., Navarro, L.M. & Martins, I.S. (2012) Global Table S4 Allometric relationships.
biodiversity change: the bad, the good, and the unknown.
Annual Review of Environment and Resources, 37, 2550. DATA ACCESSIBILITY
Pulliam, H.R. (1988) Sources, sinks, and population regula-
tion. The American Naturalist, 132, 652661. Species range maps are available at http://www.iucnredlist.
QGIS Development Team (2014) QGIS geographic informa- org/technical-documents/spatial-data. Road network data are
tion system. Open Source Geospatial Foundation Project. available at http://download.geofabrik.de/. Source references
Available at: http://www.qgis.org/en/site/ and data for life-history traits are presented in the Appendix
van der Ree, R., Smith, D.J. & Grilo, C. (2015) The ecological and in the Supporting Information.
effects of linear infrastructure and traffic. Handbook of
road ecology (ed. by R. van der Ree, D. J. Smith and C. BIOS KE TCH
Grilo), pp. 19. Wiley, Chichester.
Ana Ceia-Hasse is a PhD student working under the
Ripple, W.J., Estes, J.A., Beschta, R.L., Wilmers, C.C., Ritchie,
supervision of Henrique M. Pereira. Her work focuses
E.G., Hebblewhite, M., Berger, J., Elmhagen, B., Letnic, M.,
on process-based models of the response of biodiver-
Nelson, M.P., Schmitz, O.J., Smith, D.W., Wallach, A.D. &
sity to global environmental change. The research
Wirsing, A.J. (2014) Status and ecological effects of the
team seeks to understand the patterns and processes of
worlds largest carnivores. Science, 343, 1241484.
global biodiversity change namely the impact of
Seiler, A. & Helldin, J.O. (2006) Mortality in wildlife due to
roads on biodiversity with the aim of informing
transportation. The ecology of transportation: managing
environmental policies and ecosystem management.
mobility for the environment (ed. by J. Davenport and J.L.
Davenport), pp. 165189. Springer, Dordrecht.
Skellam, J.G. (1951) Random dispersal in theoretical popula- Editor: Marie-Josee Fortin
tions. Bulletin of Mathematical Biology, 53, 135165.
Tittensor, D.P., Walpole, M., Hill, S.L.L. et al. (2014) A mid- APPENDIX: DATA SOURCES FOR LIFE-HISTORY TRAITS
term analysis of progress toward international biodiversity USED IN THIS STUDY.
targets. Science, 346, 241244.
Blanco, J.C. (2011) Lobo Canis lupus. Enciclopedia Virtual de los Vertebrados Espa~ noles.
Torres, A., Jaeger, J.A.G. & Alonso, J.C. (2016) Assessing Museo Nacional de Ciencias Naturales, Madrid.
Jones, K.E., Bielby, J. & Cardillo, M. et al. (2009) PanTHERIA: a species-level database of life
large-scale wildlife responses to human infrastructure history, ecology, and geography of extant and recently extinct mammals. Ecology, 90, 2648.
development. Proceedings of the National Academy of Scien- Lynch, H.J. & Fagan, W.F. (2009) Survivorship curves and their impact on the estimation of
maximum population growth rates. Ecology, 90, 11161124.
ces USA, 113, 84728477. Palaz
on, S. (2010) Vison europeo Mustela lutreola. Enciclopedia Virtual de los Vertebrados
Espa~ noles. Museo Nacional de Ciencias Naturales, Madrid.
Whittington, J., St Clair, C.C. & Mercer, G. (2005) Spatial Palaz
on, S. (2012) Comadreja Mustela nivalis. Enciclopedia Virtual de los Vertebrados Espa-
responses of wolves to roads and trails in mountain valleys. oles. Museo Nacional de Ciencias Naturales, Madrid.
Pereira, H.M. & Daily, G.C. (2006) Modeling biodiversity dynamics in countryside land-
Ecological Applications, 15, 543553. scapes. Ecology, 87, 18771885.

Global Ecology and Biogeography, V


C 2017 The Authors. Global Ecology and Biogeography published

by John Wiley & Sons Ltd 9

Vous aimerez peut-être aussi